SYSTEMATIC REVIEW
Machine Learning Algorithms in Cardiology Domain: A Systematic Review
Aleksei Dudchenko1, 2, Matthias Ganzinger1, Georgy Kopanitsa2, *
Article Information
Identifiers and Pagination:
Year: 2020Volume: 13
First Page: 25
Last Page: 40
Publisher ID: TOBIOIJ-13-25
DOI: 10.2174/1875036202013010025
Article History:
Received Date: 30/10/2019Revision Received Date: 06/02/2020
Acceptance Date: 11/02/2020
Electronic publication date: 23/04/2020
Collection year: 2020
open-access license: This is an open access article distributed under the terms of the Creative Commons Attribution 4.0 International Public License (CC-BY 4.0), a copy of which is available at: (https://creativecommons.org/licenses/by/4.0/legalcode). This license permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Abstract
Background:
It could be seen in the previous decades that Machine Learning (ML) has a huge variety of possible implementations in medicine and can be of great use. Nevertheless, cardiovascular diseases cause about a third of the total global deaths.
Does ML work in the cardiology domain and what is the current progress in this regard? To answer this question, we present a systematic review aiming at 1) identifying studies where machine learning algorithms were applied in the domain of cardiology; 2) providing an overview based on the existing literature about the state-of-the-art ML algorithms applied in cardiology.
Methods:
For organizing this review, we adopted the PRISMA statement. We used PubMed as the search engine and identified the search keywords as “Machine Learning”, “Data Mining”, “Cardiology”, and “Cardiovascular” in combinations. Scientific articles and conference papers published between 2013-2017 reporting about implementations of ML algorithms in the domain of cardiology have been included in this review.
Results:
In total, 27 relevant papers were included. We examined four aspects: the aims of ML systems, the methods, datasets, and evaluation metrics. The major part of the paper was aimed at predicting the risk of mortality. A promising branch of Machine Learning, the ‘Reinforcement Learning’, was also never proposed in the observed papers. Tree-based ensembles are common and show good results, whereas deep neural networks are poorly represented. Most papers (20 of 27) have used datasets that are hardly available for other researchers, e.g. unpublished local registries. We also identified 28 different metrics for model evaluation. This variety of metrics makes it difficult to compare the results of different researches.
Conclusion:
We suppose that this systematic review will be helpful for researchers developing medical machine learning systems and for cardiology in particular.
